NDSS 2025 – L-HAWK: A Controllable Physical Adversarial Patch Against A Long-Distance Target

L-HAWK is a newly proposed controllable physical adversarial patch that can be activated by long-distance laser signals to selectively target autonomous vehicles' vision systems; the authors introduce an asynchronous learning method to optimize laser parameters and patch design, add simulation and robustness techniques, and validate the approach in digital and physical tests achieving a 91.9% average success rate at 50 meters, markedly outperforming prior work.